org.elasticsearch.indices.flush.FlushIT#testFlushOnInactive would
sometimes fail in the following case:
* SHARD_MEMORY_INTERVAL_TIME_SETTING is set very low, e.g., 10ms
* The regularly scheduled multiple flushes proceed to
  org.elasticsearch.index.shard.IndexShard#flushOnIdle
* There, the first flush will handle e.g., the first document
  that was indexed. The second flush will arrive shortly after,
  before the first flush finishes.
* The second flush will find that wasActive = true (due to the
  indexing of the remaining documents), and will set it to false.
* However, the second flush will not be executed because
  waitIfOngoing = false, and there is the ongoing first flush.
* No other flush is scheduled (since any next regularly scheduled
  flush will find wasActive = false), which creates the problem.

Solution: if a flush request does not happen, revert active flag,
so that a next flush request can happen.

Fixes elastic#87888

And fix some PR review feedback

Co-authored-by: Henning Andersen <33268011+henningandersen@users.noreply.github.com>

…ush-on-inactive

Fix some javadoc